| The current global ecosystem biodiversity is showing a serious decline.Because the interaction between species constitutes a complex food web,the changes of some species or some structures in the food web are likely to affect the stability of the food web.Therefore,accurately identifying which species or structures in the food web have significant impact on the stability of the food web will help us to determine the priority of food web protection and the priority of species protection in the process of ecosystem maintenance,so that the limited resources can be made the most efficient use.In this study,the food web from the Catlins river is taken as the research object,and the adjacency matrix is used to represent the function relationship in the food web.The relationship between the node and the robustness of the connection is deleted by cluster analysis and simulation.At the same time,the food web is proposed.The key kinds of judgment methods;secondly,using the collected global 358 food webs as research objects,using Python and R language as the operating environment,to perform descriptive statistical analysis on the food web structure indicators;finally,establish the Lasso regression model and machine respectively Learn the model,analyze the influencing factors of connection robustness and predict the connection robustness.The main results are as follows:(1)The frequency of the number of nodes,the connection density and the number of directional connections in the network structure indicators of 358 food webs are concentrated in the middle of several grouping ranges,and the value of the first grouping frequency is small,and the connection density distribution follows the power law distribution law;average The distance frequency and the diameter frequency have a similar distribution,the frequency of the first grouping range is the maximum,and the frequency of the remaining groupings is not much different;the frequency of the number of connections is similar to the frequency distribution of the centrality of the boundary,and it decreases gradually by group;the frequency of the three centrality The centrality of the intermediary in the distribution is less than the centrality of the whole,and the centrality of the whole is less than the tight centrality;most of the existing terrestrial and freshwater food webs have been studied in the food web.(2)The removal of key species has a great impact on the connection robustness of the food web.Different removal methods will also have different effects on the connection robustness.Under the random removal method,when the number of removed nodes increases,the connection robustness decreases first and then increases.In the descending order removal mode,the connection robustness drops first and then gradually increases;in the ascending order removal mode,the connection robustness basically decreases slowly,and then sharply decreases.In the random removal method,the connection robustness will have a minimum value,and the number of nodes corresponding to the minimum value is about 1/4 of the total number of nodes.Therefore,we propose the 1/4 node hypothesis,that is,when the number of species is 1/4 of the total number of species The food web has the lowest connection robustness.(3)Lasso regression model and regression tree,random forest,artificial neural network,integrated learning and other machine learning model results show that the number of nodes,number of connections,connection density,average distance,intermediary centrality,tight centrality and boundary centrality are connected to each other.The robustness has a greater impact.The rms error of the integrated learning model among the five models is the smallest and the R2 is the largest,indicating that the model has the best fitting effect,and the stacking integrated learning model can be used to predict the connection robustness.This study provides important ideas for the judgment of key species,the judgment of factors affecting robustness and the prediction of robustness. |
PDF Full Text Request